Mentor, a Siemens business, has successfully scaled its analog, RF and mixed-signal (AMS) verification software to run across 10,000 cores on the Microsoft Azure cloud platform with near-linear performance for workloads such as library characterization and verification.
The scaling optimization is a response to the need among custom-IC design teams to characterize libraries across more than 100 process, voltage and temperature (PVT) corners. Conventionally, these runs would be performed by parallelizing the work across several hundred processor cores, but this results in runtimes measured in days. By deploying across 10,000 cores, the same characterization tasks can be completed in just a few hours, Mentor claimed. Microsoft Azure demonstrated the viability of the approach using Mentor’s Analog FastSPICE (AFS) and Eldo circuit simulators.
“Mentor is pleased to continue our collaboration with Microsoft Azure to address peak library characterization and AMS verification workloads by leveraging world-class cloud platforms,” said Ravi Subramanian, general manager and vice president of Mentor’s IC verification solutions division. “Our customers have shown us the importance of utilizing burst capacity for peak demand to meet today’s design schedules, and our work with Azure has proven that we can jointly provide a robust, secure and highly scalable platform tailored to address this critical customer requirement.”
Mentor said it is continuing to work closely with cloud partners such as Microsoft Azure to help streamline the use of thousands of processors for short periods of time in AMS verification and other areas.